What We Miss When We Study Public Trust with Surveys Alone

Developers of AI data centers have a public trust problem. According to multiple surveys, a majority of Americans say they would oppose the construction of an AI data center in their local area. Across the U.S., residents express concerns about data centers’ impacts on the environment and home energy costs, noise pollution, and reduced property values–especially for residents who live in close proximity to data center campuses.  

Developers tell the public not to worry. They argue that AI data centers will have major economic benefits for communities, including job creation and increased tax revenue for municipalities to use for community improvements. And anyway, they say, most of the complaints from residents who live near existing data centers are just exaggerations.

But Americans aren’t buying it. Only 27% agree that AI data centers will have economic benefits and create jobs for their communities. Public optimism about AI data centers is low, and community groups across the country are calling for government regulations and community benefits agreements that are legally enforceable if (or when) a data center negatively impacts local residents.

AI data center developers have made a lot of promises to local communities, but those promises are ringing hollow. The question is: Why?

Survey methods: useful, yet limited, for public trust research

What we know about Americans’ overwhelmingly negative perspectives on AI data centers is largely based on survey data. Survey methods are useful for investigating public trust on a variety of issues. They can help us quantify public perspectives, measure degrees of public trust, and identify which demographic groups tend to be more trustful (or distrustful) than others. Surveys are also useful for making data-driven generalizations about a research population, which can be helpful for understanding trends in public opinion. 

However, survey data tell us only a small part of the story about public trust–whether on AI data centers or any trust-related issue. That’s because trust is too complex to be fully understood by numbers alone. 

To trust means that you’re willing to depend on, or be vulnerable to, another person or thing. That vulnerability exists because no matter how well you know someone or something, there’s no way to have 100% certainty that they will do the thing you’re expecting them to do in a way that’s favorable to you.

That’s why the act of trusting involves a kind of leap of faith. Taking that “leap” isn’t completely rational, but doing so is meaningful for our relationships with others and even our relationships with ourselves. The cultural norms and values, personal experiences, and circumstances that factor into a “trust leap” require a lot of unpacking that even the best survey questions aren’t capable of doing.

At the Center for Media Innovation and Social Impact, we emphasize the role of qualitative inquiry in our research on public trust issues around AI data centers in New England. In addition to a survey, we use semi-structured interviews and digital ethnographic observations to explore how people learn to trust (or distrust) information and information sources on data center proposals in their communities.

We recognize that to capture the complexities of trust–from its formation to its social meaning, maintenance, erosion, and repair–we need qualitative methods. Without them, trust research is incomplete.

What is qualitative research?

Qualitative researchers typically study how individuals or groups make sense of the social world. That usually means collecting non-numerical data through methods like interviews, focus groups, content analysis, or observations. For a case like AI data centers, we would ask people to explain their perspectives or analyze their online conversations about data centers in their community. A qualitative researcher might also observe communities as they organize against a local data center development. 

Qualitative methods are rarely used to measure anything or make generalizations about a research population. Instead, the overarching goal of qualitative research is to understand the social world from the perspectives of the people living in it.

In other words: survey data can tell us what people’s perspectives are, but qualitative data give us access to why they hold those perspectives.

Understanding trust dynamics

One way to understand why communities are resistant to AI data centers is to explore trust dynamics.

Exploring trust dynamics can mean focusing on several different things. “Dynamic” implies that trust involves relationships–between two people, a person and an institution, or a person and technology.

While a survey can tell you how much (or how little) people say they trust AI, it can’t tell you much about their relationship to AI (or technology, in general). Through qualitative interviews or focus groups, we can learn about how someone has incorporated AI technologies into their daily lives, how they learned to use them, or what kinds of things they’re hearing or reading that could inform their relationship to AI. We can also learn about the limits of someone’s trust by asking about the kinds of tasks they’d trust AI to complete–like writing an email–versus tasks they wouldn’t trust AI to complete–like an invasive medical procedure.

Second, when we say that trust is “dynamic,” we’re also referring to the ways that it can change over time.

Survey data tend to make trust seem like a static state–an attitude you either have or don’t have–because they provide a snapshot of respondents’ perspectives at the time when they filled out the survey. But qualitative data help us see that trust is a process; it’s something that develops over time, which means that it can also change over time, and it requires maintenance of some sort or it can erode.  

Interactive qualitative methods–like interviews or focus groups–are good tools for understanding trust as a process because we can use them to explore people’s stories about their trust relations at different points in time. And, if their trust changed in some way, we have the flexibility to ask them about events or circumstances in their lives that may have affected those trust relations.

Qualitative research puts trust in context

Another reality that surveys can obscure is that trust isn’t interpreted in the same way across all social settings. In some contexts, trusting someone may be viewed as a rational thing to do, while in other contexts trust is considered naïve. That’s because the meaning of trust is constructed in relation to things like cultural norms and values, history with institutions, or personal circumstances. Acknowledging that trust is context-dependent means that we can’t fully understand it if we assume that it’s a universal experience.

Qualitative methods are useful for exploring how people define and develop trust in context, as well as the experiences that make their trust relations meaningful.

Because qualitative research is focused on the why behind people’s perspectives, it’s great for solving contextual puzzles. For instance, state officials might be confused about why residents from two industrial towns in the same state have different reactions to new data center proposals. Without an in-depth comparison of both contexts, in which people living in those towns explain their thinking, we may never fully understand the difference.  

Another major contextual factor that shapes trust relations is a person or community’s relationship to institutions.

In some contexts, those who do not trust certain institutions base their mistrust on personal, negative experiences. Others’ mistrust comes from stories passed down by relatives or community members who have had negative experiences with those institutions. In the latter case, people socially ‘inherit’ mistrust of institutions even if they haven’t had negative experiences themselves. Qualitative methods are excellent tools for analyzing the social inheritance of trust relations because those impactful stories are often woven into the non-numerical data that qualitative researchers collect.

Finally, different contexts have different stakes for trust.

Because trust involves a leap of faith, people assume a certain amount of risk when they take a trust leap. Sometimes, the opportunities that can come from trusting someone outweigh the possible risks.

A person’s assessment of the opportunities versus the risks of trust depend on a multitude of contextual factors that are often difficult to quantify. Plus, these risk assessments are not always rational or even conscious calculations. Risks that might seem insignificant to a researcher could be surprisingly influential for a community member.

By honing in on the justifications people use for their trust (or lack thereof), as well as the history and present circumstances of a particular context, qualitative research helps us understand the contextual factors behind trust relations 

Qualitative methods matter for public trust formation–and repair

While surveys are still useful for understanding public trust in broad terms, qualitative methods are also needed to understand the nuances of trust relations. Those nuances come from the dynamic qualities of trust, as well as a multitude of contextual factors that shape trust relations. These factors are critical to understand if the goal is to build and maintain public trust.

Researchers are only beginning to understand the public backlash to AI data centers, but one thing is clear: knowing that public trust is low is not the same as understanding why.

Qualitative research helps explain the relationships and experiences that shape declining public trust, giving us a better understanding of what meaningful responses might look like. The more we know about when and why trust breaks down in a given context–from the perspectives of the people who actually experience that trust breakdown–the more that policymakers and institutions can build responses to public trust crises that actually resonate with communities.